Theses and Dissertations

ORCID

https://orcid.org/0000-0001-9490-9335

Advisor

Dash, Padmanava

Committee Member

Skarke, Adam

Committee Member

Calderon, Vladimir

Committee Member

Parajuli, Prem

Date of Degree

5-15-2026

Original embargo terms

Visible MSU Only 2 Years

Document Type

Dissertation - Campus Access Only

Major

Earth and Atmospheric Sciences

Degree Name

Doctor of Philosophy (Ph.D.)

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

This dissertation develops an integrated coastal observing and modeling framework that combines high-frequency autonomous surface vessel (ASV) observations, satellite remote sensing with machine learning (ML) retrievals, a coupled SWAT–EFDC⁺ hydrodynamic and material transport model, and imaging flow cytobot (IFCB)–based deep learning for phytoplankton community analysis. Repeated ASV transects collected measurements of chlorophyll a (Chla), colored dissolved organic matter (CDOM), turbidity, phycocyanin, phycoerythrin, pCO₂, salinity, temperature, pH, and dissolved oxygen from 2021 to 2024 and were analyzed using Seasonal Trend decomposition, Mann–Kendall tests, principal component analysis, hierarchical clustering, and generalized additive models. These observations revealed strong seasonal, spatial, and interannual variability across the Western Mississippi Sound. ML models trained with Landsat-derived remote sensing reflectance and ASV matchups produced high skill retrievals of key optically active parameters. Extreme Gradient Boosting achieved R² = 0.957 (MAE = 0.226 µg/L) for Chla and R² = 0.978 (MAE = 0.708 ppb) for CDOM, while Random Forest performed best for turbidity (R² = 0.965, MAE = 0.220 NTU). These models enabled reconstruction of 2014–2024, 30 m resolution time series, capturing winter–summer Chla peaks and spring maxima in CDOM and turbidity linked to riverine inflows. A coupled SWAT–EFDC⁺ model developed for 2023–2024 reproduced hydrodynamics with high skill, including water surface elevation (R² = 0.85–0.90; RMSE ≈ 0.05 m), salinity (R² up to 0.99; RMSE 0.34–3.53 ppt), and temperature (RMSE 0.12–0.39 °C; R² ≥ 0.996). River discharge comparisons showed strong agreement, especially for the Pearl River (R² = 0.97). Simulated material transport exhibited clear spatial structure, including strong southward export through Transect 2 (>1×10⁸ units·m³/s), northward inflow through Transects 5 and 7 (>6×10⁷ units·m³/s), and peak eastward transport through Transect 5 (~2×10⁸ units·m³/s). An IFCB–deep learning pipeline classified 57 phytoplankton taxa (H′ = 3.02; J′ = 0.71) with 91.1% accuracy and identified toxin-forming genera. Community shifts were primarily associated with turbidity, temperature, pH, and salinity. Together, these components deliver decision‑ready, spatially explicit products that enhance long‑term environmental monitoring and provide a transferable framework for predictive, multimodal assessment of estuarine systems under rising climatic and anthropogenic pressures.

Sponsorship (Optional)

This research was supported by federal funding from NASA EPSCoR (Award No. G00007006); the Mississippi Department of Environmental Quality and the U.S. Department of the Treasury through the Resources and Ecosystems Sustainability, Tourist Opportunities, and Revived Economies of the Gulf Coast States Act of 2012 (RESTORE Act) (Award Nos. 8006490-02.01 MSU and 8007148-02.01 MSU); and the U.S. Army Engineer Research and Development Center, Vicksburg, MS, USA (Award No. W912HZ-19-2-0019). The statements, findings, conclusions, and recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of NASA EPSCoR, the U.S. Army Engineer Research and Development Center, the Mississippi Department of Environmental Quality, or the U.S. Department of the Treasury.

Download

Share

COinS